It’s been a quiet summer for YASP, but we’re back with some new features, just in time for TI!
Stats are useful because you can use them to predict the future. We’ve taken a look at all the pro games in 6.88 to give you the best chance of getting your predictions right! Check out our insights here.
A new tab on each player page now shows the games that that player has played with professional players! The list is from a Steam API endpoint called GetProPlayerList.
Denies @ 10
We’ve updated our ranking system to more accurately rank the best players at the top.
We’ve increased weighting of MMR again for the 2016Q3 ranking cycle, which started 2016-07-01. Essentially, a player with half the MMR of another player now needs 64 times as many wins to have the same ranking score.
We’re proud to announce our new partnership with DotaCoach to help you improve your game. DotaCoach.org is an on demand service sort of like an “Uber” for learning Dota. Review your YASP stats with a coach or ask them any questions you have on how to improve. Or just play a game together. Interacting with a skilled player is one of the best ways to improve at the game.
We love partnerships like these that help us further our mission: to make you a better Dota player.
And as always, we’re always improving our infrastructure to improve site reliability and performance.
After a lengthy migration, we’ve moved over 1.2 billion matches from PostgreSQL to a 4-node Cassandra cluster. We’re seeing over 98% of requests served in under a second! Yay! This is a huge improvement over the performance we were seeing with PostgreSQL. Now, load times scale linearly with the number of matches processed, so player pages with lots of matches (~5000+) might have slightly slower times, but this is still much better than the 15-30 second times from before!
Native support for compression has also cut our disk space usage in half! We were even able to switch to HDD storage for a 4x reduction in storage costs. All this means future savings.
We moved Postgres, Redis, and each Cassandra node to their own separate compute instances to prevent them from contending for memory, which had previously been causing random crashes.